check ninode and nfile by command

[Chen Yingjie]

Can someone advise me how to check ninode and nfile by command? not the SAM.

Thank you.

[federico_3]

use :


sysdef | grep -e ninode -e nfile

federico

[federico_3]

You can use als":

echo "ninode/D"|adb -k /stand/vmunix /dev/kmem

echo "nfile/D"|adb -k /stand/vmunix /dev/kmem

regards,
federico

[Carlos Fernandez Riera]

with sar -v 5 5 you will see :
number inodes / configured inodes
number opened file / number of files

[Vladislav Demidov]

Try to perform the next procedure:
1. Create a sar daily data directory
mkdir /var/adm/sa/sa`date +%d`
2. run the sar command
sar -v 1
refer to sar man page for additional information

[CHRIS_ANORUO]

Use sar -v 5 5 => This will show ninode as inod-sz (current size and set value) Same with nfile file-sz. You can also querry the running kernel with the adb command as foolows: echo "ninode/D"|adb -k /stand/vmunix /dev/kmem

[James R. Ferguson]

Hi:

Glance's table option (toggled with 't') will show this.

...JRF...

[Alan Riggs]

I think that should be enough options for you (I use sar -v personally). Just a reminder: the ninode limit is a hard limit only for hfs filesystems. vxfs filesystems will dynamically allocate more inodes when needed.

[Wodisch]

Hello,
I do not agree with Alan, "ninode" has nothing to about how many inodes may
exist on a file system. It is just a cache (not a table, like "nfile") for used inodes.
Hence it is not possible to guess/check, if it is "full" - even if all "slots" in that
cache are used, usually some are just occupied by already closed inodes, making
access to them faster - if they are used again. Use the mentioned "sar" option to
test for overflows! And for the "nfile" - that is a table, "sar" is correct about it, but
if you are about to experience trouble you could "reserve" some entries in there
for super-user processes, by increasing the kernel parameter "file_pad".
HTH,
Wodisch

[Wodisch]

Hello,
I do not agree with Alan, "ninode" has nothing to about how many inodes may
exist on a file system. It is just a cache (not a table, like "nfile") for used inodes.
Hence it is not possible to guess/check, if it is "full" - even if all "slots" in that
cache are used, usually some are just occupied by already closed inodes, making
access to them faster - if they are used again. Use the mentioned "sar" option to
test for overflows! And for the "nfile" - that is a table, "sar" is correct about it, but
if you are about to experience trouble you could "reserve" some entries in there
for super-user processes, by increasing the kernel parameter "file_pad".
HTH,
Wodisch

[Wodisch]

Hello,
I do not agree with Alan, "ninode" has nothing to about how many inodes may
exist on a file system. It is just a cache (not a table, like "nfile") for used inodes.
Hence it is not possible to guess/check, if it is "full" - even if all "slots" in that
cache are used, usually some are just occupied by already closed inodes, making
access to them faster - if they are used again. Use the mentioned "sar" option to
test for overflows! And for the "nfile" - that is a table, "sar" is correct about it, but
if you are about to experience trouble you could "reserve" some entries in there
for super-user processes, by increasing the kernel parameter "file_pad".
HTH,
Wodisch

[Wodisch]

Hello,
I do not agree with Alan, "ninode" has nothing to about how many inodes may
exist on a file system. It is just a cache (not a table, like "nfile") for used inodes.
Hence it is not possible to guess/check, if it is "full" - even if all "slots" in that
cache are used, usually some are just occupied by already closed inodes, making
access to them faster - if they are used again. Use the mentioned "sar" option to
test for overflows! And for the "nfile" - that is a table, "sar" is correct about it, but
if you are about to experience trouble you could "reserve" some entries in there
for super-user processes, by increasing the kernel parameter "file_pad".
HTH,
Wodisch

[Alan Riggs]

Wodisch, I think you may have misinterpreted what I wrote. Perhaps I was not clear.

I was saying that the ninode results reported by sar are not a meaningful measure of any limit on a vxfs filesystem.

[Alan Riggs]

Wodisch, I think you may have misinterpreted what I wrote. Perhaps I was not clear.

I was saying that the ninode results reported by sar are not a meaningful measure of any limit on a vxfs filesystem.